A surrogate-based framework for time-dependent reliability-based topology optimization under multi-source uncertainties

Considering the time- dependent uncertainty of random or degradation processes in topology optimization is crucial for achieving efficient and reliable structural design over the entire service life. However, the rigorous quantification and evaluation of these uncertainties often result in a substantial increase in computational burden. To improve the efficiency of structural design under multi-source uncertainties, this study proposes a quantile-based active learning framework for time-dependent reliability-based topology optimization (T-RBTO). First, stochastic processes are employed to characterize time-dependent uncertainties in loads and material properties. Subsequently, two different learning functions are used to obtain the extreme response, and the Kriging model is updated based on the quantile. By introducing quantiles, the traditional reliability constraints are transformed into quantile constraints, significantly improving the computational efficiency of reliability constraints. Finally, the optimized parameters that satisfy time-dependent reliability constraints are incorporated into the topology optimization process to achieve lightweight structural designs. Several benchmark topology optimization cases involving time-dependent uncertainties, such as Gaussian process loads and material degradation processes, demonstrate the accuracy and computational efficiency of the proposed T-RBTO framework.